Re: 1.1.1 Branch notice 1500 ET (that's 3:00 PM ET for the non geeky)
Most of the remaining issues marked for 1.1.1 seem to have been fixed for 1.1.1 but are still open pending different solutions for 1.2. There is one big exception -- an ERROR during shutdown on Java 5. It would be great if someone could look at that before the branch. I've been trying hard to claim that Geronimo works fine under Java 5, and this is a little unfortunate. http://issues.apache.org/jira/browse/GERONIMO-2134 Thanks, Aaron On 8/2/06, Matt Hogstrom [EMAIL PROTECTED] wrote: The memory leak is fixed. Thanks to Kevan for the quick work. I am going to branch this afternoon at 1500. If you have something in progress ping me and we'll work it out. The branch will be branches/1.1.1 and will only be taking bug fixes and performance regressions (tracked by JIRA.). All existing JIRA's that were not resolved in 1.1.1 will be moved to 1.x. If there is interested in a 1.1.2 speak up. Thanks all.
Re: 1.1.1 Branch notice 1500 ET (that's 3:00 PM ET for the non geeky)
On Aug 2, 2006, at 10:39 AM, Aaron Mulder wrote: Most of the remaining issues marked for 1.1.1 seem to have been fixed for 1.1.1 but are still open pending different solutions for 1.2. There is one big exception -- an ERROR during shutdown on Java 5. It would be great if someone could look at that before the branch. I've been trying hard to claim that Geronimo works fine under Java 5, and this is a little unfortunate. http://issues.apache.org/jira/browse/GERONIMO-2134 Agreed that an ERROR isn't very desirable. I don't know anything about the origins of this code, but it seems to be a workaround for a JRE memory leak of Class objects. ConfigurationClassLoader and MultiparentClassLoader are using introspection to clear out SoftCache's in ObjectInputStream and ObjectOutputStream. It would seem that in Aaron's Java 5 JRE, ObjectInputStream doesn't contain a field named subclassAudits. While Joe's 1.5 implementation does contain the field. Hopefully, this mean's the memory leak has been fixed in Aaron's version... Best fix I can think of at the moment is to reduce the log severity to info or debug. Comments? --kevan On 8/2/06, Matt Hogstrom [EMAIL PROTECTED] wrote: The memory leak is fixed. Thanks to Kevan for the quick work. I am going to branch this afternoon at 1500. If you have something in progress ping me and we'll work it out. The branch will be branches/1.1.1 and will only be taking bug fixes and performance regressions (tracked by JIRA.). All existing JIRA's that were not resolved in 1.1.1 will be moved to 1.x. If there is interested in a 1.1.2 speak up. Thanks all.
Re: 1.1.1 Branch notice 1500 ET (that's 3:00 PM ET for the non geeky)
I just tried on a fresh install of Geronimo-Jetty-J2EE-1.1 on Linux (SuSE 10.1) with JDK 1.5.0_07 (the one SuSE distributes) and I get the ERROR. I don't know how to tell if my JVM has a class leak error or not, but I'll try to check if you know. In any case, eliminating the message is fine for now. Thanks, Aaron On 8/2/06, Kevan Miller [EMAIL PROTECTED] wrote: On Aug 2, 2006, at 10:39 AM, Aaron Mulder wrote: Most of the remaining issues marked for 1.1.1 seem to have been fixed for 1.1.1 but are still open pending different solutions for 1.2. There is one big exception -- an ERROR during shutdown on Java 5. It would be great if someone could look at that before the branch. I've been trying hard to claim that Geronimo works fine under Java 5, and this is a little unfortunate. http://issues.apache.org/jira/browse/GERONIMO-2134 Agreed that an ERROR isn't very desirable. I don't know anything about the origins of this code, but it seems to be a workaround for a JRE memory leak of Class objects. ConfigurationClassLoader and MultiparentClassLoader are using introspection to clear out SoftCache's in ObjectInputStream and ObjectOutputStream. It would seem that in Aaron's Java 5 JRE, ObjectInputStream doesn't contain a field named subclassAudits. While Joe's 1.5 implementation does contain the field. Hopefully, this mean's the memory leak has been fixed in Aaron's version... Best fix I can think of at the moment is to reduce the log severity to info or debug. Comments? --kevan On 8/2/06, Matt Hogstrom [EMAIL PROTECTED] wrote: The memory leak is fixed. Thanks to Kevan for the quick work. I am going to branch this afternoon at 1500. If you have something in progress ping me and we'll work it out. The branch will be branches/1.1.1 and will only be taking bug fixes and performance regressions (tracked by JIRA.). All existing JIRA's that were not resolved in 1.1.1 will be moved to 1.x. If there is interested in a 1.1.2 speak up. Thanks all.
Re: 1.1.1 Branch notice 1500 ET (that's 3:00 PM ET for the non geeky)
Matt Hogstrom wrote: The memory leak is fixed. Thanks to Kevan for the quick work. I am going to branch this afternoon at 1500. If you have something in progress ping me and we'll work it out. The branch will be branches/1.1.1 and will only be taking bug fixes and performance regressions (tracked by JIRA.). All existing JIRA's that were not resolved in 1.1.1 will be moved to 1.x. If there is interested in a 1.1.2 speak up. Thanks all. Given that 1.2 is not going out until winter, I'm going to fix some stuff for 1.1.2. I've already created a version in Jira. Regards, Alan
